We manage your security to weather the AI storm →

← Resource Hub

Which Bridge Has Never Been Hacked? Track Records Compared

By Daniel GelfandPublished 12 min read

Which Bridge Has Never Been Hacked? Track Records Compared

Quick answer. No bridge can prove it was never hacked. Through November 14, 2026, we found no protocol-level exploit for Chainlink CCIP, Across, deBridge or Circle's CCTP contracts. Wormhole, Ronin and Nomad were exploited, as were integrations on LayerZero, Axelar and CCTP.

Watch: Which Bridge Has Never Been Hacked? Track Records Compared 1:46
Video transcript

No bridge can prove it was never hacked. A clean record is often a treasury's first question, and the easiest claim to misread.

Through November 14, 2026, we found no protocol-level exploit for CCIP, Across, deBridge, Stargate or Circle's CCTP contracts. Each record rests partly on the provider's own claims.

CCIP's current design, CCIP 2.0, went live on September 28, 2026, so its record under that design is short.

Integrations fail too. On August 19, 2026, Allbridge's CCTP router lost about 190 thousand dollars by crediting an attested message that never burned USDC.

Wormhole, Ronin and Nomad were exploited in 2022. In 2026, a Kelp DAO route on LayerZero lost about 292 million dollars.

Detection matters. Ronin's theft was found six days later; Secret says Axelar told it seven days after. Wormhole paid a 10 million dollar bounty for a near miss.

In Guardian's Olympus LayerZero report, a bridge meant to pin two DVNs encoded optional DVNs as zero, meaning use the default. A later default change could raise fees or block delivery.

Before bridging, read the live verifier set onchain. Ask about independent RPC sources, source authentication, rate limits, upgrade timelocks and monitoring.

Read the full article for both tables. Shipping a bridge or bridged token? Guardian's smart contract audit puts two independent teams on one scope.

Why This Question Matters

A clean record is often the first thing a treasury asks for. It is also the easiest claim to misread.

2026 sharpened the question. Kelp DAO's rsETH route on LayerZero lost 116,500 rsETH, about $292M, on April 18. An Axelar integration with Secret Network lost about $4.67M on June 10. LayerZero, Secret Network

This page asks what a clean record proves. Our ranking of the biggest bridge hacks covers the losses.

How We Built the Table

Never hacked here means no publicly documented protocol-level exploit in the sources we reviewed, up to the date this page was updated. No source can prove a negative.

The label does not cover front ends, staff, relayers, integrations, token contracts or connected chains. Any of those can fail while the core protocol holds, and several cases below did.

The tables cover bridges secured by their own verifiers, signers or relayers.

Canonical rollup bridges, such as Arbitrum's, OP Stack's and Base's, rest on rollup proofs and upgrade keys, so they are out of scope.

Out of scope is not clean: in September 2022, Offchain Labs paid 400 ETH for an unexploited Arbitrum bridge bug. Decrypt

Hop, Synapse, Celer cBridge, Hyperlane and Polygon PoS are not covered here.

We list pauses and near misses next to exploits. A route that halts for days can still strand treasury funds.

Volumes are each provider's own figures, and they measure different things: transfer volume, transferred value or settled volume. Use them for scale, not for ranking. Chainlink Foundation, LayerZero, deBridge

Bridges With No Exploit on Record

These systems had no protocol-level exploit in the sources we reviewed, though an integration built on CCTP was exploited. Each record rests partly on the provider's own claims.

CCIP's docs describe a 16-operator committee that needs 9 attestations, source-chain finality, issuer-set rate limits and timelocked security-critical upgrades. Our LayerZero vs Chainlink CCIP comparison sets that committee against app-chosen verifiers. Chainlink Documentation, Chainlink

Its mainnet record starts July 17, 2023, the shortest of the five. Its current design, CCIP 2.0 with optional Cross-Chain Verifiers, went live on September 28, 2026, so its record under that design is short.

Across says it has never lost user funds to a protocol-level failure across $35B+ in volume. Across

In 2025, a compiler mismatch in a Lens upgrade paused Lens bridging for three to four days. Across called it an operational mistake, with no user funds at risk.

On November 10, 2026, an attacker took across.to through social engineering at its registrar, Namecheap. Across recovered it and promised to make users whole; see our Across domain hijack report. Across Protocol

Circle launched CCTP on Ethereum and Avalanche on April 26, 2023. It moves native USDC by Circle-attested burn and mint. Circle

We found no exploit of Circle's CCTP contracts. But Allbridge's CCTP router on Base lost about $190K on August 19, 2026, because it credited an attested message that never burned USDC. SlowMist, Odaily

In August 2024, Asymmetric Research disclosed a bug in Circle's noble-cctp that skipped the TokenMessenger sender check. It was fixed with no funds lost. Asymmetric Research

ProtocolVerification modelLive sinceValue moved (own figure)Pauses and near misses
Chainlink CCIP16-operator committee; 9 must attestJuly 17, 2023; CCIP 2.0 since Sept 28, 2026$26.34B cumulative, October 2026None found in reviewed sources
AcrossRelayers front funds; UMA optimistic settlementNovember 8, 2021$35B+, May 2026Lens upgrade pause, 2025; domain hijack, November 2026
deBridgeElected validators; 8 signatures, two-thirds of the setFebruary 17, 2022$23B+ settledStaff phishing attempt, August 2022
Circle CCTPCircle attestation; native USDC burn and mintApril 26, 2023No primary figure foundAllbridge integration exploit, Aug 2026; Noble sender-check bug, fixed 2024
StargateLayerZero messagingMarch 17, 2022No primary figure foundGovernance phishing post, Dec 2023, about $43K from users

Bridges That Were Hacked

Five well-known systems have a documented exploit. Three failed in the protocol's own code or keys; two failed on a route or integration.

Scale did not prevent loss, as the volume column shows.

Two cases sit at the protocol's edge. LayerZero says route security is the application's to configure, and Secret says Axelar's core protocol was not affected. Our Kelp DAO breakdown sets out LayerZero's and Kelp's accounts. LayerZero, Secret Network

So a protocol's record and a route's record can differ. Ask which one any clean-record claim describes.

Ronin shows why detection belongs in the comparison. Our Ronin bridge hack breakdown covers how the keys were taken.

Axelar shows it too. Secret Network says Axelar notified it on June 17, seven days after the June 10 exploit, and still says not to bridge via Axelar.

Wormhole also had a near miss: on February 24, 2022, a whitehat reported an uninitialized proxy bug in its Ethereum core bridge. Wormhole paid a $10M bounty, and no funds were lost. Immunefi

SystemVerification modelLive since, volumeIncidentAfterward
WormholeMessages need 13 of 19 Guardian signatures (VAAs)Early 2021; $40B+ by Sept 2024Feb 2, 2022: 120,000 wETH, about $320M to $326M; Solana signature check bypassedService halted; 1:1 backing restored
LayerZero (Kelp DAO route)App-chosen DVNs per routeBeta by March 2022; $270B+ by June 2026Apr 18, 2026: 116,500 rsETH, about $292M; sole DVN fed compromised RPC dataLayerZero Labs refuses sole-DVN channels
Axelar (Secret integration)Validator network; 75 active validators in 2025Jan 13, 2022; $10B+ by June 2025Jun 10, 2026: about $4.67M unbacked tokens; counterfeit-chain deposit bypassed source checksSecret told June 17; Axelar bridging still disabled
Ronin bridgeFive of nine validator signatures (2022 design)Ronin mainnet, February 2021Mar 23, 2022: about $624M (DefiLlama); five validator keys compromisedFound six days later; legacy bridge moved to CCIP, 2025
NomadMessages checked against a trusted rootNo launch date verifiedAug 1, 2022: about $190M; upgrade made messages valid by defaultFigure is gross, before any recovery

Why Track Record Is Not Enough

A clean record shows that no attack has succeeded so far. It does not show which attacks a design would survive.

The failures above fall into four classes: compromised keys, flawed message checks, unsafe configuration, and upgrade or infrastructure mistakes. Guardian Resource Hub

Upgrades reset part of any record. Nomad's loss followed one, and so did Across's 2025 Lens pause. Across

Headline verifier counts can mislead too. A route's resolved settings can look right today while its raw configuration still inherits defaults that can change. Guardian

Guardian found this in Olympus's LayerZero integration. The bridge meant to pin exactly two required DVNs, but encoded optional DVNs as 0, which means use the default (M-01, Medium, resolved).

A later LayerZero default change could then add a DVN requirement without governance action, raising fees or blocking delivery.

The fix used LayerZero's NIL sentinel, type(uint8).max, for no optional DVNs. Reading raw app config, not only resolved config, is what shows it.

For the mechanics behind each failure class, see our guide to cross-chain bridge security.

What to Check Before Bridging

Ask for evidence on the exact route and chain pair you will use. Then confirm it onchain, yourself or through a reviewer.

For LayerZero routes, our LayerZero DVN configuration guide lists the onchain calls that show live settings.

Teams shipping a bridge or bridged token can scope the review with our cross-chain bridge audit guide.

CheckEvidence to ask forRed flag
Verifier set and thresholdLive verifier list and threshold for your route, read onchainOne verifier, or settings left on defaults
Operator independenceWho runs each verifier, signer and nodeSeveral keys or verifiers under one company
RPC diversityIndependent data sources per verifier, and quorum rulesOne provider, or no rule when sources disagree
Source authenticationHow the destination checks source chain and senderDeposits accepted from unverified chains or senders
Rate limitsInbound and outbound caps per route and time windowNo caps, or caps sized to total supply
Pause controlsWho can pause a route, and how fastOne key both pauses and unpauses
Supply reconciliationBacking against minted supply on every chainNo public reconciliation or gap alerts
Upgrade controlsTimelock length and who can bypass itInstant upgrades from a single key
Incident disclosureWritten reports for every pause and incidentClean-record claims with no incident log
MonitoringBlock-by-block alerts on outflows and configuration changesUsers report problems before the team does

Frequently Asked Questions

Can a bridge that was hacked be a better choice than one that was not?

It can. A record does not show controls added since. After the Kelp exploit, LayerZero Labs began requiring a quorum of independent RPC sources for its DVN. Compare the current controls on your route, not past headlines. LayerZero

Was 2026 a quiet year for bridge hacks?

No. Guardian's DefiLlama-based ranking counts 51 flagged bridge hacks worth $753.5M in 2026 through November 5, against $38.3M in all of 2025. Both are gross figures, before recoveries. Guardian Resource Hub

How often should a treasury recheck a bridge route?

Before each large transfer, and after any upgrade, verifier change, pause or incident on that route. Verifier sets, contract versions and volumes all change, so last quarter's review may describe a different route.

Sources

Check the route before value moves through it

Shipping a bridge or bridged token? Guardian's smart contract audit puts two independent teams on one scope, covering OFTs, DVNs, peers, shared decimals, replay and compose flows.